Hi Google Beam👋! What started as Project Starline has evolved into a revolutionary 3D video communications platform. The combination of our AI video model and light field display allows you to make eye contact and read subtle cues as if you were face-to-face 🤯

A consistent theme we’ll talk about today is the idea of taking cutting-edge research and making it a reality. Here are a few examples of research efforts making their way into new and existing products 🧵

It's simple math. Livestreaming today using a single camera data stream is already pushing the limits of today's internet. What do you think will happen when 6 cameras attempt to stream simultaneously? Streaming six camera feeds simultaneously will significantly strain current internet infrastructure, especially for high-resolution video like 4K or 8K. Here’s a breakdown of what’s likely to happen: Bandwidth Overload: A single 4K stream at 30fps typically requires 15-25 Mbps with efficient compression (e.g., H.265). Six streams could demand 90-150 Mbps or more, exceeding many residential and even commercial internet plans. Most households in the US average 100-500 Mbps download but far less upload (10-50 Mbps), which is the bottleneck for livestreaming. Latency and Buffering: Without robust upload speeds and optimized encoding, latency will spike. Real-time streaming relies on low-latency protocols like WebRTC or SRT, but multiplexing six streams increases packet loss risks, leading to buffering or degraded quality on the viewer’s end. Compression and Processing: To manage bandwidth, platforms might aggressively compress streams, reducing quality (e.g., dropping to 1080p or introducing artifacts). Encoding six feeds in real-time also requires powerful local hardware (high-end CPUs/GPUs) or cloud-based encoding, adding cost and complexity. Network Congestion: If multiple users in a region stream six-camera setups, ISPs could face congestion, especially in areas with outdated infrastructure. This could trigger throttling or prioritization issues, particularly for non-commercial users. Viewer-Side Challenges: Viewers need sufficient download bandwidth and devices capable of decoding multiple streams. Multi-view interfaces (e.g., split-screen) could overwhelm lower-end devices, causing lag or crashes. Potential Mitigations Adaptive Bitrate Streaming: Dynamically adjusting resolution/bitrate based on network conditions could help but may compromise quality. Edge Computing: Offloading encoding to cloud servers near the streamer could reduce local hardware demands. Next-Gen Protocols: Emerging standards like QUIC or 5G’s low-latency capabilities might alleviate bottlenecks, but 5G adoption and coverage remain uneven. Content Delivery Networks (CDNs): Robust CDNs can optimize viewer delivery, but upload to the CDN remains a hurdle. Future Outlook Widespread six-camera streaming will require significant upgrades: broader fiber-optic access, 5G/6G proliferation, and advancements in compression (e.g., AV1 or future codecs). Until then, only users with high-end setups (e.g., fiber connections, dedicated encoding rigs) or platforms with heavy server-side optimization will manage it effectively. Expect early adopters to face inconsistent performance, with quality trade-offs common. If you’re planning this, what’s your setup (camera resolution, internet speed, platform)? I can tailor suggestions. Assessing streaming limits The request is about streaming six cameras at once, and it's tough since one stream already maxes out internet. I'm curious how this'll pan out with multiple streams. Currently looking for info on bandwidth needs for multiple HD streams and potential bottlenecks. This'll help figure out the challenges involved.

Discover the future of AI investing. AIS delivers exposure to the companies driving the next wave of innovation—semiconductors, data centers, and AI applications. Explore the supercycle today.

Just so we're clear, I came up with the idea of an array of cameras that create a composite of the user to facilitate eye contact during video calls back in 2007. I just didn't know how to build it. But thanks for the credit.

soon

Not good